How To Assemble A Desktop PC/Overclocking

Overclocking is the practice of making a component run at a higher clock speed than the manufacturer's specification. The idea is to increase performance for free or to exceed current performance limits, but this may come at the cost of stability.

Overclocking is like supeing up a car, if you just want to get where you're going, there's no need for it. But it is fun and educational and can get you a machine that provides performance all out of proportion to its cost.

Think of the 3GHz on your new 3GHz Pentium 4 as a speed limit asking to be broken. Some other components in your computer can also be overclocked, including RAM and your video card in many cases. Over clocking is possible because of the way electronic parts, especially VLSI (Very Large Scale Integration) chips are made and sold. All processors in a given line, the Pentium 4 for example, are made the same way, on a large die that is cut up into individual processors, those processors are then tested and graded as to speed, the best chips will be marked as 3.0 GHz the second best 2.8 etc. As time goes by and production processes and masks improve, even the lower rated chips may be capable of faster speeds, especially if vigorous cooling is implemented. Also many manufacturers will mark chips that test faster at slower speeds if there is higher demand for the lower end component.

It’s important to note that not every chip will be overclockable; it’s really the luck of the draw. Some companies that sell ‘factory overclocked’ systems engage in a practice called “binning” where they buy a number of processors, test them for overclocking potential and throw the ones that don’t overclock in a bin to be resold at their rated speed. Even with processors that have a reputation for overclocking well, some parts simply will not exceed their rating.

That said, effective cooling can give a boost to a chip's overclockability. With luck you will be able to get extra performance out of your components for free. With luck and skill you can get performance that is not possible even when using the top of the line components. Sometimes you can buy cheaper parts, and then OC them to the clock speed of the higher end component, though the cost of extra cooling can compromise any money you may be saving on the part, not to mention warranty and part life issues.

WARNING: OVERCLOCKING MAY VOID THE WARRANTY ON THE PARTS BEING OVERCLOCKED. MAKE SURE YOU KNOW WHAT YOU ARE DOING, AS OVERCLOCKING MAY ALSO CAUSE SYSTEM INSTABILITY, AND MAY ALSO CAUSE DAMAGE TO COMPONENTS AND DATA. REMEMBER THE 3 "C'S" WHEN OVERCLOCKING: CAREFUL, CONSERVATIVE, and CAUTIOUS.

Things that can't be overclockedEdit

Although it is possible to overclock many of the components of a computer (such as the CPU, FSB frequency and video card), it is not possible to overclock all components. For example, it is not possible to overclock a hard disk drive nor an optical drive such as a CD-ROM drive. For such devices other solutions exist, i.a. use a quicker file system and/or faster components in the first place; also in some special cases of hard disks drives, update the driver (which contains optimized code); remap blocks with high latencies to low latency blocks (using a tool such as mhdd), and in event a RAID is present, change the configuration, driver and/or software and/or settings. However such techniques and procedures are beyond the scope of this document.

ComponentsEdit

CPUEdit

The CPU's clock speed is the FSB (Front Side Bus) clock speed (base, not effective speed) times the CPU's multiplier. On most newer CPUs (notable exceptions: AMD Black editions, AMD Athlon FX series and Intel Extreme Editions), the multiplier is locked, so you will have to adjust the FSB clock speed (However, it might be possible to 'unlock' the chip's multiplier on some older chips. See CPU Locking.) The FSB is not adjustable on some motherboards, and almost all OEM systems. The FSB and multiplier, if not locked, are adjustable from within the BIOS. Note that upping the FSB clock speed also increases the clock speed of many other components, including RAM.

When increasing the FSB clock speed, do so in small increments of a few MHz at a time. After you do this, boot up your computer to make sure it works. If your computer successfully boots, increase the FSB some more. If it won't boot, lower the FSB until your computer properly boots up. Repeat until you have the highest setting with which your computer will boot up. Next, test your OS to make sure it is stable by running a burn application, or any application that uses 100% CPU power. If a crash or reboot results, lower the FSB speed some more until it runs smoothly. On some motherboards you are also able to change the voltage of the CPU and other components in order to help stabilize the system. However, this increases the components' heat output and can harm or shorten the life of your system instead.

Notes: On AMD K8-based chips (Athlon 64, Opteron, Turion, and Socket 754 and 939 Semprons), there is no FSB- there is an integrated memory controller (IMC) and a HyperTransport bus (HTT). The IMC has a base clock speed like a FSB does and for overclocking, it would be adjusted just like a FSB would. The HTT can also be overclocked like the CPU core can be. Its base clock speed is the exact same as the IMC's and thus by default you will overclock the HTT bus as you overclock the CPU. Note that the HTT bus has a multiplier and it is adjustable. Many motherboards do not function well with the HTT bus frequency much over the stock frequency- that's where the adjustable HTT multiplier comes in. (It should be noted that overclocked systems are most stable when the HTT is at or below 1000Mhz)

Also, some newer desktop processors (AMD K8 series with Cool 'n Quiet, Intel Pentium 4 6xx series, Pentium D 830, 840, stepping C1 Pentium D 9xx series with Enhanced Intel Speed Step) and most mobile processors (exc. Celeron M) can vary their operating frequency while running by lowering the CPU multiplier. This results in multipliers below the highest one being unlocked. This allows for a very high FSB/base clock speed with a lower multiplier to use very fast RAM to its full potential without overclocking the CPU so much that it is not stable.

How to choose the best CPU for overclockingEdit

Assuming that you have selected a quality motherboard, high-quality RAM , thermal solution and power supply; you may wonder why your processor won't exceed a certain speed limit. Lets assume that you have a memory chip that is capable of taking the maximum frequency the motherboard can throw at it and yet, when you exceed a certain speed limit you realize that your system becomes unstable.

A PCI bus generally runs at 33 Mhz. When you exceed 35-36 Mhz, the hard disk and other IDE devices become unstable, because the IDE controller is controlled through the PCI bus. Oftentimes, you may encounter texture corruption, when your AGP bus exceeds a certain speed limit. This was often observed on older motherboards that wouldn't allow you to lock the AGP and the PCI bus at stock speeds.

The good news: regardless of the FSB speed, most motherboards nowadays automatically ensure that the frequency of the PCI, AGP and other buses always remain constant (in other words; their speeds are locked unless you deliberately change them). This implies that the other components connected to the motherboard don't undergo stress when the FSB speed is raised. You have the fastest RAM you could find and the obscenely fast speed ratings on your ensure that you can extract the most performance by bumping up your FSB speed to the limits.

The main culprit that plays the spoilsport is your CPU. Even if you have an exceptionally good thermal solution, your CPU won't exceed a certain limit.

Example: I had experimented with a Pentium III 700E Mhz processor and a Pentium III 800E MHz processor on an Asus CUBX-E motherboard using Kingston PC-133 SD-RAM. The reason I chose these 2 processors for experimentation was because they both used a FSB speed of 100MHz. This motherboard was really flexible, I was able to increment the FSB to 150Mhz. I was able to extract 1050 MHz from the stock 700Mhz. This is because the multiplier is 7, which unfortunately cannot be changed. So I bumped up the FSB from 100Mhz to 150 Mhz; which gave me the resultant speed of:

7
(Multiplier)
× 150
(Front Side Bus)
= 1050 MHz
(resultant frequency)

Simple arithmetic? Yes. Now, logically speaking, if I can extract 1050Mhz from a 700Mhz processor; I should be able to extract 1200Mhz from an 800 MHz processor. This is not true. I tried doing exactly the same with the 800 MHz processor and the Computer crashed. However, it was stable when I set the FSB speed to 133 MHz. When I set the FSB at 133 MHz ; this was the result:

8
(Multiplier)
× 133
(Front Side Bus)
= 1064 MHz
(resultant frequency)

This simple experiment shows that a CPU gets saturated after a certain clock speed. Typical symptoms of an erratic CPU include instability and at times, you may not be able to boot up at all.

This particular CPU die was manufactured using a 0.18u process. When Intel launched a similar CPU using a 0.13u process; they shipped those CPUs with the stock speed of up to 1.4Ghz. This CPU core was based on the P6 Architecture and it used a 10 stage pipeline. Presently, Intel manufactures the Pentium-M CPU which is based on the P-6 architecture; the difference being that they manufacture it using a 0.09u process and they have increased the depth of the pipeline.

These terms may seem cryptic and this concept may be difficult for some to grasp. It's really very simple: To be a successful overclocker; you need to purchase the best CPU possible; not necessarily the fastest. Always go for a processor that uses the latest manufacturing process. A CPU rated at 3Ghz which is manufactured by using a 0.13u process won't overclock as well as a CPU that is rated at 2.6Ghz using a 0.09u process.

Deeper pipelines ensure that the CPU has the capability to scale higher in terms of speed. The disadvantage is that a CPU with a deeper pipeline is slower than a CPU that uses a smaller pipeline assuming that they are running at the same speed. AMD Athlon CPUs are famous for their relatively short pipelines. Thats why they perform better than the Pentium 4 CPUs at the same clock speed. Before purchasing the fastest processor, always keep this in mind. Choosing a processor smartly helps you extract the maximum speed out of your machine. You don't need to know what a pipeline exactly does. Refer to the processor spec sheet, find out these basic details of the CPU core and its architecture and choose accordingly.

To increase the computer's stability, you may also disable the spread spectrum; set the PCI speed to 100MHz; set the voltage to the middle range, not too high; and disable any smart fan settings. and those settings work for all recommended boards.

Video cardEdit

Two different parts of a video card may be overclocked[1], the GPU (Graphics Processing Unit) and the RAM. In addition, disabled pipelines on a video card may also be enabled through third-party drivers, third-party software, or direct hardware modifications depending on your video card type. Overclocking a video card is usually done through third-party or proprietary software.

Recent ATI proprietary Catalyst drivers feature an interface called Overdrive that allows for dynamic GPU frequency scaling based on its temperature and load. Increase the load, the clock rate increases for performance, but it's balanced against the increasing temperature. Sufficient for simple increases in overall performance, this doesn't allow for the best performance increase which requires overclocking the memory. For that you'll need third-party applications or drivers.

An application example is ATITool. This program has many options, including GPU and memory overclocking, temperature monitoring, and fan control, allowing for a much more complete solution to overclocking ATI based video cards. There are many third party drivers, omegadrivers.net, for example hosts ATI and nVidia drivers as well. Both of which include integrated overclocking and many unlocked features, including enhanced image quality for nVidia-based cards.

nVidia video cards can also be OCed through a hidden feature in the driver called coolbits. Coolbits is a feature that can be unlocked by creating a DWORD in regedit for windows operating system. To use the coolbits feature, for windows just simply open the regedit then open the directory HKEY_LOCAL_MACHINE>Software>NVIDIA Corporation>NVTweak and create a new DWORD value in the NVTweak folder named coolbits then right-click it>modify type 3 for single card or 1A for SLI in the value name .It is a good a overclocking tool as it has a fairly conservative "optimal clock" once you have thus increased the core clock (not the memory clock!!) run a gpu intensive task like 3dmark, repeat until you have a sudden drop in the benchmark score. This is the thermal throttling kicking in; do not push it any harder as it will result in permanent damage to your gpu. Back off the clockspeed by about 20-30mhz

The most important thing to remember about overclocking a video card is cooling. This can't be stressed enough. Just as a CPU can be damaged or have a shortened lifespan by overclocking or excessive and prolonged heat, so can a video card. In the past year many inexpensive and easy to install options have surfaced for cooling a video card, from adhesive ram heatsinks which attach to un-cooled ram chips, to rather expensive water-cooling solutions. A good midpoint (both in cost and effectiveness) solution is to purchase and install a direct exhaust, "sandwich" cooling solution. Direct exhaust means all air from the cooling fan is blown across the video card and directly out of the computer case, usually using the open PCI slot below the AGP (or PCIe) slot. This allows for substantially lower GPU temperatures.

A sandwich cooler is two aluminum or copper heatsinks, shape formed for a particular video card, that "sandwiches" the video card in between the two and are usually connected by some kind of copper heat pipe which allows for the hotter side to convey heat to the cooler side for dissipation. The GPU should never surpass 60 degrees Celsius for optimal performance and to avoid damaging the card. Most of the latest video cards are rated to go up to 90c, but this is NOT recommended by anyone. The optimal temperature for a video card is 40-55c for the card itself (the GPU's temperature differs depending on which you have,) but the lower you can get it, the better.

It is also possible to use software to change the fan speed on certain cards. Changing the fan to run at full speed can cool the card better, dependening on your card and the speed of the fan in the first place. Software such as Rivatuner can be used for Nvidia based cards.

One important thing to note. Many think that the option which says "AGP voltage" in their BIOS can be used to "voltmod" a video card to get a bit more power out of it. In fact, it's used for something else, and raising the AGP voltage can and probably will cause damage to a video card.

Getting the few extra MHz out of a chipEdit

CoolingEdit

When increasing the speed of any computer components you are making the components work harder and by doing so they output more heat. Heat can cause system instability so cooling is necessary to help keep your components stable at higher speeds. Without good cooling you could harm or shorten the life of your system. CPU temperature can usually be checked from within the BIOS. However, these are inaccurate as your CPU is under almost no load in the bios. SiSoftware Sandra may be used within Windows to check temperature. This should be done when your CPU has been under a heavy load for a while for optimum results.

There are three types of cooling that are generally accepted for overclocking: Air, water, and peltier.

With both air cooling and water cooling some type of transfer material is needed to move the energy away from the sensitive electronics. The device used for this purpose is a heatsink. The two most popular heatsink materials are Aluminum and Copper. The heatsink that is stock on factory computers by major manufacturers (Dell, Gateway, IBM) is usually made of aluminum, which has satisfactory heat transfer characteristics. However when overclocking more heat is being produced from the increase in power consumption. In order to obtain lower temperatures a material with better heat transfer properties is important and copper is the material that offers the best ratio of price/performance.

PowerEdit

Chips at higher speeds may need more power. Raising the vcore voltage on a CPU might enable it to go at slightly faster speeds but by doing so you add a lot more heat output from the CPU. The vcore of a processor is the voltage at which a chip is set to run at with the stock speed. This voltage may need to be changed when the multiplier is raised because otherwise the transistors in the chip won't switch fast enough - transistors switch faster the higher the supply voltage. If there is not enough voltage then the chip will begin to make mistakes and give bad data results. Good cooling is needed to keep the system stable at higher speeds. Raising the vcore too much may harm or shorten the life of your system. Raising the vcore can also greatly affect the stability of the system. This is where a high quality PSU will come into play. While many cheap, no-name brand PSU's will crash and die with more vcore, a good quality one will live to serve you for a long time.

Note: increasing the speed (multiplier or fsb) without changing the voltage will also increase heat output, but not as much as when also increasing voltage. Having said that, increasing the multiplier or FSB without adjusting the voltage may make your system unstable (undervolt).

ReferencesEdit

  1. Coles, Olin. "Overclocking the NVIDIA GeForce Video Card". http://benchmarkreviews.com/index.php?option=com_content&task=view&id=205&Itemid=38. Retrieved 2008-09-05. 

External linksEdit


Choosing and installing the software · Silencing

Last modified on 13 June 2013, at 15:46